Is it possibly to have $L_\infty$ bounds on the eigenfunctions of the Laplacian operator on bounded regular domains with Dirichlet condition? I found several papers by Sogge but these are pretty general and apply to Laplacian-Beltrami operator on manifolds. Their paper has bounds that are possibly too loose for my application.
Thanks, John
Moser iteration proceeds roughly like this: If the dimension $n$ is greater than $2$ and we assume homogeneous Dirichlet , we can proceed as follows: Using the Sobolev inequality on $\mathbb{R}^n$, $$ c(n)\|u\|_{2n/(n-2)} \le \|\nabla u\|_2 $$ and integrating by parts, we get something roughly like this (you have to figure out what to do with all the absolute values) for each $p > 1$: $$ \lambda\int |u|^p \ge \int |u|^{p-1}(-\Delta |u|) = \frac{4(p-1)}{p^2}\int |\nabla |u|^{p/2}|^2 \ge \frac{4(p-1)}{p^2}c(n)\|u\|_{np/(n-2)}^p. $$ Therefore, given $p_0 > 1$, if we set $p_k = p_0(n/(n-2))^k$, we get $$ \|u\|_{p_{k+1}} \le \left(\frac{p_k^2}{4(p_k-1)}\frac{\lambda}{c(n)}\right)^{1/p_k}\|u\|_{p_k}. $$ Iterating this, we get $$ \|u\|_\infty \le C(n,p_0)\lambda^{n/(2p_0)}\|u\|_{p_0}. $$ The power of the eigenvalue in the final estimate doesn't need to be calculated explicitly. You know what it has to be by observing that the left side is invariant under rescaling space ($\mathbb{R}^n$) and therefore the right side must be, too.
it is possible to bound the $L^\infty$ norm of an eigenfunction in terms of its $L^2$ norm, a power of the relevant eigenvalue and a dimensional constant. this is true for every domain of $R^N$ with finite measure, no matter how bad the boundary is (I am assuming homogeneous Dirichlet boundary conditions, of course). I don't know if this answer your question.
